[CVE-2026-98866][MODERATE 7.0] sched/cache: Introduce task_struct->sched_cache_grp to fix UAF [ Upstream

[CVE-2026-98866][MODERATE 7.0] sched/cache: Introduce task_struct->sched_cache_grp to fix UAF [ Upstream

[AL-KERNEL]

CVE: CVE-2026-98866
Priority: MODERATE 7.0
AL-KERNEL base severity: MODERATE
KPANIC flag: YES
Patch: sched/cache: Introduce task_struct->sched_cache_grp to fix UAF [ Upstream
Commit: 9436130e6da72af63289a4acb3aa0c1df07a3928
Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=9436130e6da72af63289a4acb3aa0c1df07a3928
Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98866
Analysis date: Fri, 23 Oct 2026 21:59:14 -0400
ActionableScore: 5
ActionableScore lower bound: unknown
Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review
Manual review required: YES

Summary:
A local unprivileged process can race repeated fork exec or exit transitions with scheduler cache accounting, causing an active `sched_cache_group` use after free with fixed field reads and writes that can crash the kernel and may corrupt a reclaimed slab object.

======================================================================
ABOUT THIS REPORT
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98866 is available here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98866

The original announcement does not normally provide a security severity
estimate, CVSS assessment, or enough information to determine whether the
reported kernel bug represents a practically relevant security issue.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin. It combines
an autonomous classifier with LLM-assisted technical analysis and a
separate ActionableScore mechanism.

The purpose of this report is to prioritize Linux kernel CVEs before
manual review, identify cases that require prompt investigation, and
support automatic closure of issues that are unlikely to have meaningful
security impact.

Published priority for this report: MODERATE 7.0
Manual review required: YES

A detailed explanation of the methodology and priority rules is included
at the end of this message.

======================================================================
AL-KERNEL CLASSIFICATION RESULT
======================================================================

CVE-2026-98866	MODERATE	https://www.kernelcve.org/list/?q=CVE-2026-98866 CHECK WITH IMPACT FROM ORIG NN MODERATE | AEGIS IMPACT MODERATE_KPANIC AEGIS_AS=6 LEGACY=MODERATE_KPANIC RESULT=SAME (final reconciliation agrees; DECISION: both MODERATE_KPANIC)	Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:H';*CWE-416;CWE-362;*CWE-664;Other CVSS 'AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H';BEST CVSS score: '5.8';DESCR 'The scheduler accessed a sched_cache_group through task mm while the task address space could be replaced or released by exec or exit processing. The mm reference and its sched_cache_group reference did not provide a lifetime guarantee suitable for scheduler hot path access, allowing account_mm_sched and related fair scheduler paths to reach a group after its final reference was dropped. A local user can exercise the relevant lifetime transitions by creating tasks and repeatedly invoking fork exec and exit while scheduler accounting runs concurrently. CONFIG_SCHED_CACHE is required, and paths involving scheduler cache accounting and possibly NUMA activity must execute. The race is local rather than network reachable and requires timing between scheduler activity and mm replacement or teardown. The established primitive is a use after free involving reads and accounting updates to fields of sched_cache_group. This is memory corruption, but the available patch text does not establish attacker control over object reuse, an arbitrary address write, or reliable privilege escalation. Impact is at least local denial of service and may include limited confidentiality or integrity impact if the freed allocation is reused before the stale scheduler access. For the CVSS the PR:L is used for the highest score paranoid vector because an unprivileged local user must execute processes and invoke local process lifecycle operations to create the race. AC:H reflects the required concurrency and timing. The paranoid CIA:L,L,H assessment recognizes the confirmed use after free and possible fixed field corruption without claiming a stronger arbitrary write or privilege escalation primitive. Manual review is required because this is a scheduler use after free with a write capable accounting path and the supplied source is not sufficient to rule out controllable reclaim or broader corruption. Automated impact decision path. The original SimpleNet classifier selected MODERATE_KPANIC. The primary LLM estimated paranoid CVSS 5.8 and the production ActionableScore returned 5 with recommendation Actionable Moderate / Strong Important candidate requiring manual review. The production rules then evaluated CVSS thresholds and ActionableScore thresholds. Recorded rule transitions were CVSS and initial triage stage MODERATE_KPANIC to IMPORTANT because the CVSS score, vector or bug class threshold rule was applied. KP and FD specialist checks stage IMPORTANT to MODERATE_KPANIC because KP specialist did not establish a qualifying panic. The legacy KP specialist produced a review artifact and its decision was processed by the applicable KP rules. The legacy FD or DH downgrade specialist has no review artifact for this run. Exact triggered rule markers were INCREASED_TO_HIGH_BASED_ON_GUESSCVSS and DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFKP. The independent weaker Aegis shadow evaluated CVSS 2.5 and ActionableScore 6 with primitive class CLASS_B and manual review REQUIRED. Its independently reconciled impact was MODERATE_KPANIC. The shadow agreed with the final production impact and aggregation did not change that impact. The final reported impact is MODERATE_KPANIC.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 5)	MAYBE	READ DANGER DISK INIT UAF KPANIC  KPANIC INCREASED_TO_HIGH_BASED_ON_GUESSCVSS DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFKP	-	-	checked

======================================================================
ACTIONABLESCORE ANALYSIS
======================================================================

ActionableScore=5
PrimitiveClass=CLASS_B
RealUAF=YES
HostAdminRequired=NO
NamespaceScopedPrivilege=NO
ImportantCandidate=YES
ManualReview=REQUIRED
AutoDowngradeAllowed=NO
RequiresHostUID0=NO
AuthorizationDecisionBypass=NO
ProtectedResourceImpact=NONE
B4PendingReview=NO
B4PendingConfirmed=NO

1. ActionableScore

* Raw conservative score: 4
* Conservative score: 5
* Raw paranoid score: 5
* Paranoid score: 5
* The real UAF floor 5 is applied because an ordinary local process can repeatedly drive fork exec and exit transitions while the stale scheduler object is used for active accounting rather than cleanup only.
* Conservative CVSS: CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H (4.7)
* Paranoid CVSS: CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:H (5.8)
* CWE: CWE-416;CWE-362;CWE-664
* Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**::

2. Signal breakdown

* Plus 2. A real use after free is explicitly established. Scheduler code can reach `sched_cache_group` through `p->mm` after the group lifetime is no longer protected.
* Plus 1. The trigger uses ordinary process lifecycle operations including fork exec and exit. These operations do not require host administrator privileges.
* Plus 1. The stale object is used by active scheduler accounting. The visible accounting logic reads `grp->footprint` and writes an updated value rather than only dereferencing the object during cleanup.
* Plus 1 in the paranoid calculation. A fixed field write into a reclaimed slab allocation creates plausible limited integrity impact. This does not establish an arbitrary write or controlled replacement.
* FLOOR 5 is applied to the conservative score. The attacker can repeatedly drive lifecycle transitions, and the stale use performs nontrivial active read and write processing.
* No hard race penalty is applied. The bug is a race, but the supplied evidence does not establish a narrow one shot or externally uncontrollable timing window. The relevant lifecycle operations can be repeated.
* No remote reachability point is awarded. No network trigger is established.
* No confidentiality point is independently awarded in the actionable arithmetic. The stale read is internal, and no disclosure path to userspace or another observer is shown.
* No strong corruption point is awarded. Controlled reclaim, attacker selected replacement contents, arbitrary read or write, callback control, and privilege escalation are not demonstrated.
* No extra availability point is awarded for the same UAF crash consequence.
* No repeated release or refcount point is awarded. The bug involves missing lifetime protection and stale access, not a demonstrated second put or release through the same dangling reference.
* Paranoid repeated refcount leverage is withheld because there is no concrete second lifetime mutation.
* B5-R is not applicable. There is no retry accumulated OOB write.
* B3-PENDING is not applicable. There is no shared backing write.
* B4-PENDING is not applicable. There is no authorization DENY to ALLOW transition.

3. Reachability analysis

The credible attacker is an ordinary local user who can create processes and invoke fork exec and exit repeatedly. No host UID 0 requirement or capability check is visible for these lifecycle operations. The scheduler independently executes `account_mm_sched` and related fair scheduler accounting while a task mm is being replaced or detached.

The vulnerable kernel must be built with CONFIG_SCHED_CACHE. The specific footprint accounting shown in the supplied context also depends on the applicable scheduler cache and NUMA accounting paths being active. These are feature prerequisites, not evidence of a separate host privileged setup step.

The complete trigger is local and does not depend on a namespace capability. NamespaceScopedPrivilege is NO because no namespace privilege is needed, not because a namespace grants the operation. There is no network, device, firmware, or special hardware trigger.

The attacker can repeatedly create the relevant process lifecycle transitions. Successful collision with scheduler accounting still depends on concurrency, so CVSS AC:H remains appropriate. The available evidence does not prove deterministic reproduction, but it also does not support a separate hard race penalty in the actionable score.

For the CVSS the PR:L is selected because the attacker must have a local user context and execute process lifecycle syscalls. PR:N would incorrectly treat local syscall access as requiring no user context. Host administrator powers are not required for the established trigger.

4. Primitive and severity interpretation

The root cause is a lifetime mismatch between scheduler access and ownership of `mm->sched_cache_grp`. Scheduler hot paths obtained the group through a task mm, but replacement or teardown of that mm could release the group while scheduler accounting still retained an access path through the task. The task itself did not own a reference that remained valid for the scheduler access.

The relevant call sequence is scheduler fair accounting reaching `account_mm_sched`, followed by access to the group obtained through `p->mm->sched_cache_grp`, while exec or exit changes the task mm and releases the mm owned group reference. The stale operation can read and update accounting fields such as `footprint`.

This is an active structured object UAF, not fixed async stale storage and not cleanup only. The stale object participates in scheduler calculations and receives a fixed field accounting write. This supports PrimitiveClass CLASS_B and RealUAF YES.

The demonstrated primitive is a stale object read and fixed field write. A kernel crash or other availability failure is credible. Limited integrity impact is defensible in the paranoid CVSS because the write can occur after slab reuse. Limited confidentiality impact is retained only in the paranoid CVSS because stale structured data can be read, but no user visible disclosure is established.

The evidence does not establish controlled reclaim, attacker controlled replacement contents, an arbitrary target, function pointer use, credentials modification, or reliable local privilege escalation. CIA:H,H,H is therefore not defensible. The patch and supplied discussion establish a UAF but not a complete exploitation chain.

ImportantCandidate is YES because this is a write capable active scheduler object UAF reachable below host administrator privilege, and the consequences of slab reuse remain unresolved. AutoDowngradeAllowed is NO because automatic crash only classification would discard a concrete active post lifetime write primitive. This does not prove Important severity. The supported operational minimum is Actionable Moderate with mandatory exploitation leverage review.

The fix gives every task its own reference to `sched_cache_group`. The reference is acquired during fork and exec, released during exit, and scheduler code accesses `p->sched_cache_grp` rather than relying on the changing mm lifetime.

No authorization decision bypass, protected resource access, information disclosure path, or public privilege escalation evidence is supplied. The cited reports corroborate the lifetime defect, but no stronger advisory or exploit consequence is present in the provided material.

5. One-sentence report phrase

A local unprivileged process can race repeated fork exec or exit transitions with scheduler cache accounting, causing an active `sched_cache_group` use after free with fixed field reads and writes that can crash the kernel and may corrupt a reclaimed slab object.

6. Manual review recommendation

MANUAL CHECK REQUIRED

The bug is a confirmed active object UAF with a write capable scheduler accounting path and below host administrator reachability. Controlled reclaim and privilege escalation are not established, but slab reuse consequences cannot be safely reduced to crash only without reviewing the complete `cache_sched.c` and `fair.c` call paths, allocation cache, RCU grace periods, and reproducibility.

7. Call-site confidence

Call-site confidence: medium

======================================================================
UPSTREAM PATCH SUMMARY
======================================================================

Patch: sched/cache: Introduce task_struct->sched_cache_grp to fix UAF [ Upstream
Commit: 9436130e6da72af63289a4acb3aa0c1df07a3928
Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=9436130e6da72af63289a4acb3aa0c1df07a3928

Changed files:
  kernel/sched/cache_sched.c
  fs/exec.c
  kernel/fork.c
  kernel/exit.c
  include/linux/sched.h

Diff excerpt:

Not included in this email. See the upstream URL for the full patch.

Full patch:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=9436130e6da72af63289a4acb3aa0c1df07a3928

======================================================================
DETAILED REPORT METHODOLOGY
======================================================================

The original Linux kernel CVE announcement for CVE-2026-98866 can be found here:
https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-98866

The original CVE announcement normally does not include a security-level
estimate. In particular, it may not contain a CVSS assessment, an impact
level, or enough information to determine whether the reported bug is a
practically relevant security issue. One purpose of this parallel CVE list
is to provide that missing technical and prioritization information.

The original goal of the AL-KERNEL project was to prioritize Linux kernel
CVE analysis automatically before manual review. The system can also help
identify non-security issues that may be suitable for automatic closure.

This report was generated by AL-KERNEL, an AI-assisted Linux kernel
vulnerability analysis system developed by Alexander Larkin.

The first analysis stage combines an autonomous classifier with additional
LLM-based analysis. The autonomous classifier runs locally on a CPU and is
based on a backpropagation neural network. Together, these mechanisms
produce a technical vulnerability description, identify likely weakness
types, estimate CVSS severity, and provide input for ActionableScore.

Two CVSS estimates are retained because incomplete kernel vulnerability
information often permits more than one defensible interpretation:

  Conservative CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H
  The Best / paranoid CVSS vector: AV:L/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:H
  The Best / paranoid CVSS score: 5.8

The conservative vector represents a lower-impact interpretation.
The Best/paranoid vector intentionally represents a plausible upper-bound
interpretation and should not automatically be treated as demonstrated
real-world impact.

CVSS may also need to be adjusted for a particular Linux deployment,
because actual reachability, privileges, enabled kernel configuration,
hardware, namespaces, exposed device nodes, and other environmental
conditions can differ significantly between systems.

A separate ActionableScore mechanism evaluates practical remediation
urgency. Its analysis may include reachability, attack prerequisites,
subsystem exposure, memory-corruption characteristics, denial-of-service
reliability, and possible confidentiality, integrity, or
privilege-escalation impact.

  Conservative ActionableScore: unknown
  Paranoid ActionableScore: 5

The final base severity is taken directly from the second tab-separated
field of the AL-KERNEL classification result. ActionableScore does not
replace or independently override that final AL-KERNEL decision, and
if ActionableScore adjusted impact level of ALKERNEL, then you would see
self-readable flags above like INCREASED_TO_HIGH_BASED_ON_ACTIONABLESCOREHIGHEREQTHAN7.

For an AL-KERNEL result of MODERATE, this report uses the following
additional presentation split:

  ActionableScore below 5   -> MODERATE REGULAR
  ActionableScore 5 or more -> MODERATE 7.0

The distinction between MODERATE REGULAR and MODERATE 7.0 makes it
possible to identify Moderate issues that should receive manual analysis
and fixes before lower-priority MODERATE REGULAR issues. In many cases,
MODERATE REGULAR fixes may wait for a later rebase or routine update.

There is one override in which MODERATE REGULAR becomes MODERATE 7.0
even when the ActionableScore is below 5. When the AL-KERNEL result
contains the KPANIC flag, a MODERATE result is always presented as
MODERATE 7.0. The KPANIC flag selected with few regexps without
usage of AI at all, so it helps to detect cases when Kernel Crash happens
and similar (to filter False-Negative results from the LLM usage).

KPANIC indicates that a reliable kernel crash, kernel panic, or similarly
serious kernel availability impact was identified by the classification
workflow.

AL-KERNEL base severity for this report: MODERATE
KPANIC detected for this report: YES
Published priority for this report (same as in Subject): MODERATE 7.0

These results are intended to support engineering triage. They are
machine-generated estimates, and cases marked for manual review should
be validated by a human security engineer before final disposition.
For more info read docs linked from here: https://kernelcve.org/
(and you can submit you own patch there to generate such a report
for non-existant CVE-id yet).

Note that in many cases this AI tool selects higher severity, than
real is (means you can expect Importants instead of Moderate 7.0 or
Moderates 7.0 instead of regular Moderates). If you see such cases,
please use reply email interface to add additional manual analyses
info to this particular CVE.
And please, please, let me know when you see Lows instead of Importants
or Important instead of Low (because particular for such cases I
need to tune this AI tool to make it better for this one and next similar).
My contact email for such notifications is alexanjelausa@gmail.com (and both
send reply to CVE record itself too and see "reply" button below for howto reply).